Journal of Gerontology 1979 Vol. 34. No. 3, 358-367

A Study of Age Categorization1 Nathan Kogan, PhD2

UMEROUS investigators have used questionnaire procedures to determine N respondents' generalized beliefs about the chronological age onsets and durations of phases of adulthood. Most of these studies have been directed to the question of when old age begins (Blau, 1956; Harris & Associates, 1975; Peters, 1971; Shanas, 1962; Tuckman & Lorge, 1953; Zola, 1962). There are two studies, however, that have examined beliefs about the onset of age stages across the adult life span (Cameron, 1969; Drevenstedt, 1976). These perceived onsets have been related to such respondent characteristics as age, sex, and social class. Although it is useful to have information about generalized beliefs regarding age stage onsets, there are certain difficulties inherent in this type of research. Harris and Associates (1975) observed that 50% of their respondents did not provide a chronological age estimate when asked to specify the age at which the average man or woman becomes old. These respondents maintained that the definition of old age depended on health, retirement, or other unspecified circumstances. Given that 50% of the sample was unwilling to provide a concrete age estimate despite the fact that the question called for it, one must seriously consider the possibility that those who do provide

'I should like to express my indebtedness and gratitude to Stephanie Brody and Kathleen Connor for serving as experimenters, Joel Kostin for assistance in data analysis, and Kimberlee Goodwin for preparation of the figures. An earlier version of the present paper was presented at the 30th Annual Scientific Meeting of the Gerontological Society, San Francisco, Nov. 1977. 2 Prof., Dept. of Psychology, Graduate Faculty, New School for Social Research, 66 West 12th St., New York, NY 10011.

358

an age figure are simply responding to the "demand character" of the question. Under these latter conditions, age estimates may have an "off-the-top-of-the-head" quality, and bear little relation to the respondents' age categorizations of specific people. The issue can be addressed empirically, of course, by providing stimulus persons for chronological age estimations and age categorizations (e.g., middle-aged, elderly). For the various age stages, then, it should prove feasible to determine the chronological ages that respondents associate with both the onset and the midpoint of these stages. No study of age categorization would be complete that did not take account of the sex of the stimulus persons. It has been claimed (Williams, 1977) that women are perceived to age more rapidly than do men. The present research examines this issue. Further, age categorizations may be influenced by the age and sex of the respondent, and hence these are also explored in the present research. Thus far, the focus has been on the cognitive component of age categorization. The use of concrete stimulus persons, however, permits an examination of the affective component in age judgments as well. It is relatively straightforward to obtain differential preference data for the stimulus persons. These are examined in relation to the age and sex of both respondents and stimulus persons. Finally, the present paper considers the matter of the overlap of age category boundaries. Taken at their face value, the studies of generalized beliefs cited earlier suggest the presence of sharply demarcated boundaries

Downloaded from http://geronj.oxfordjournals.org/ at Indiana Univ. School of Medicine on September 2, 2012

This study used photos (stimulus persons) to explore age categorization in a sample of 150 subjects of diverse ages: 15 males and 15 females in each of five age groups: 18-21, 22-28, 29-38, 39-55, and 56-76. Two sets of 33 photos each (a male and a female set) were presented to subjects for chronological age estimation, age categorization (as adolescent, young, middle-aged, elderly, and aged adult), and preference. Results indicated considerable overlapping of estimated ages across age categories, suggesting that the boundaries between adjacent age stages are highly permeable. Sex-of-photo exerted a major influence on categorization and preference. Female, relative to male, stimulus persons were assigned to older age categories, were perceived to attain middle-aged and elderly status sooner, and were younger when chosen as most preferred (but only for male subjects). These sex-bias effects were mitigated in the oldest respondents. Relevance of the present research to problems of age stereotyping is discussed.

AGE CATEGORIZATION

in the oldest group had attended or graduated from college; the remaining eight subjects had graduated from a secondary school. In sum, with the exception of a minority of the oldest group, all subjects were currently attending or had previously attended college. Materials. — The stimulus persons were depicted in achromatic passport-size photographs (6.3 cm x 6.3 cm). These were obtained from a different city than the one in which the study took place, and hence there was little likelihood that a respondent would personally recognize any of the stimulus persons. A total of 66 photographs was used, 33 of males and 33 of females. Both sets were distinguished by wide variation in the ages of the persons depicted.

Procedure. — All subjects participated on an individual basis. The examiner informed the subjects that the study concerned the perception of the age of persons represented in photographs. The male or the female set of photos was then shown to the subjects one at a time. Upon the presentation of each photo, subjects were requested to estimate the chronological age of the stimulus person depicted. After exposure of the 33rd photo, the entire set of 33 was handed to the subject with instructions to select the two stimulus persons considered to be the most similar. Following this procedure, subjects were asked to group the photos into five categories: adolescent, young adult, middle-aged, elderly, and aged. No restrictions were placed on the number of stimulus persons assigned to each category. METHOD Upon completion of this task, the subjects were asked the following question: "If you Subjects. — A total of 150 subjects partici- had an opportunity to become acquainted with pated in the study. For purposes of analysis, three of these people on the basis of the photos these were divided into 15 males and 15 females alone, which three would you choose?" The in each of five groups: 18-21, 22-28, 29-38, three persons selected were noted by the 39-55, 56-76. It should be noted that this par- examiner. ticular age clustering was made solely for the If a subject was first given the male photo purpose of equalizing ATs per cell. The age set, the entire procedure as just described classification took place prior to the analysis was repeated for the set of female photos. of the data. The youngest age group was com- Similarly, subjects who first responded to the prised of students at a nearby liberal arts col- female set were subsequently given the male lege. The remaining age groups included stu- set. The sex-of-photo sequence was alterdents in graduate and adult education pro- nated across subjects. The order of photo pregrams. In the case of the oldest group, some sentation for age estimation was randomized; subjects were recruited at a nearby senior each subject saw the photos in a different citizens center. Twenty-two of the 30 subjects random order.

Downloaded from http://geronj.oxfordjournals.org/ at Indiana Univ. School of Medicine on September 2, 2012

between age stages. A chronological age range is associated with an age stage (e.g., middleaged adult), and no age overlap is found with adjacent younger and/or older age stages. The strong possibility exists that these nonoverlapping age stages reflect subjects' preferences for clearly defined categories. It is of considerable theoretical and practical interest to find out whether these sharply defined boundaries between age stages are maintained when a range of stimulus persons is to be classified. Do respondents attribute unique age ranges to stimulus persons categorized as young, middle-aged, and elderly, for example? Or, rather, does one find overlap in the age ranges assigned to the various age stages? Whichever alternative is closer to the observed outcome, we can expect that the results will have implications for the dynamics of age stereotyping. If old age can be defined in strict chronological age terms, hence setting apart persons who have reached that age from all others, the cognitive preconditions for stereotyping have been clearly established. On the other hand, if classification as old should bear only an imperfect relation to chronological age, the criteria for class membership become multidimensional and concomitantly the conditions for stereotyping should be mitigated. It is not the intent of the present paper, however, to explore age stereotyping directly. The focus of this study is on the influence of respondent age and sex on the age categorization of male and female stimulus persons of diverse ages.

359

360

KOGAN

and middle-age, and that the provision of such a category would have reduced boundary overCategory boundaries. — The first question lap. A test of this possibility entailed a comasked of the data concerned the relative per- parison with the boundary between middlemeability of age category boundaries. Im- aged and elderly adult, where it is difficult to permeable boundaries would imply sharp de- conceive of any intervening age category. lineation of age categories on the basis of The respective percentage overlaps for the chronological age. Consider the boundary two boundary conditions were 64.0% for between middle-age and elderliness, for ex- young vs middle-aged adult and 59.4% for ample. If the boundary were impermeable, all middle-aged vs elderly adult. The difference of the stimulus persons considered elderly between these percentages is negligible. We would be older than all of the stimulus persons can legitimately conclude, therefore, that the considered middle-aged. In contrast, a per- permeability of boundaries between age catemeable boundary would imply overlap. Some gories is a genuine psychological phenomenon, of the stimulus persons assigned to the middle- rather than an artifact of the particular age aged category would be older than some as- categories employed in the study. Interpretation of the foregoing findings must signed to the elderly category. The present data offer an opportunity to assess the magni- take account of the fact that respondents first tude of this kind of boundary permeability estimated the age of the stimulus persons in the photos and subsequently grouped the photos or overlap. For the present purpose, the two extreme into age categories. There is not the slightest age categories — adolescent and aged — were doubt that a reversal of the sequence — groupdropped because a considerable number of re- ing by age categories first and estimating spondents did not use those categories at all chronological ages second — would comor used them quite sparingly. Hence, the pletely alter the pattern of the findings. Once analysis is confined to the three middle cate- categories are formed, respondents would gories of young, middle-aged, and elderly adult. inevitably make certain that chronological How permeable is the boundary between young age estimates are nonoverlapping across cateand middle-aged and between middle-aged gories in order to prove that they are capable and elderly adult? Given the use of male and of judging consistently. With such a procedure, female stimulus persons, each respondent we would obviously conclude that age category 3 had four opportunities to manifest overlap boundaries are essentially impermeable. Where then does the truth lie? It is apparent or lack of overlap in categorizing. The general finding regardless of boundary condition, that there are likely to be dual truths linked sex of photo, sex of subject, or age of subject to the method employed. If ecological validity was the presence of a substantial degree of is a major consideration, however, one should overlap. Thus, 150 subjects across four con- not choose a procedure whose demand charditions generated 600 opportunities for over- acter calls for complete consistency. Grouplap. There were in fact 370 cases in which ing by age category followed by chronological overlap occurred, a rate of 61.6%. When this age estimation clearly possesses such demand figure was broken down by respondent age character. The continuous numerical scale group, the percentages were as follows: 47.5 of chronological age can be easily chopped for 18 to 21-year-olds, 63.3 for 22 to 28-year- up to fit the age categories formed. On the olds, 55.0 for 29 to 38-year-olds, 65.8 for other hand, when age estimations are made 39 to 55-year-olds, and 76.6 for 56 to 76-year- first, the compulsion toward consistency is olds. Thus, there was a trend toward greater much reduced, for age categorization is deboundary permeability with increasing re- pendent on more than chronological age alone. spondent age. Demographic and personality factors are likely It is possible, of course, that the large magnitude of overlap observed reflects the set of age categories employed. No category was A limited empirical test of this logical inference was carried out on a sample of seven graduate psychology students using the same set of 66 photos provided between young and middle-aged employed in the experiment proper. Age categorization preceded chronadult, for example. It can be argued that there ological age estimation. Not a single case of boundary overlap (out of 28 opportunities) was found. On the average, a gap of about 5 years separated is a category of mature adult between young adjacent age categories. RESULTS

Downloaded from http://geronj.oxfordjournals.org/ at Indiana Univ. School of Medicine on September 2, 2012

3

AGE CATEGORIZATION

361

I II II II II II

Si

ll

il

Downloaded from http://geronj.oxfordjournals.org/ at Indiana Univ. School of Medicine on September 2, 2012

to play a role. Knowledge that a person is re- stimulus persons. This difference was highly tired and inactive as opposed to employed and significant, F (1, 140) = 64.01,/? < .001, to2 = active would almost certainly influence age .10. Of course, we do not know whether the categorization over and beyond knowledge observed difference is in the "eye of the of chronological age. Of course, the use of beholder," or is nothing more than a bias in photos rules out demographic information, but the selection of photos. Also worthy of note the expressive cues offered by photos could is the presence of a marginally significant very well lead to personality inferences. interaction of sex of photo with age of subThese in turn could influence age categoriza- ject, F (4, 140) = 2.69, p < .05, w2 = .02. The tion. After all, such expressions as "she looks tendency toward assigning more female than especially young for her age" or "he looks male photos to the older age categories is quite old for his age" are commonplace. In most pronounced for the youngest subjects other words, age category boundaries may be and least pronounced for the oldest respondfuzzy and permeable rather than sharp and ents. The presence of this interaction indiimpermeable in regard to chronological age cates that photo selection bias cannot entirely account for the findings. Where the female demarcations. stimulus persons are concerned, younger subjects are more inclined toward older age assignFrequency of Use of Age Categories Since subjects were free to assign as few or ments in comparison to the age assignments as many photos to each of the five age cate- of older subjects. gories as they wished, it is of interest to know how the photos were distributed. For this Chronological Age Estimations purpose, weights of 1, 2, 3, 4, and 5 were asA powerful age-of-photo effect was obsigned to the categories from adolescence tained,^ (2, 280) = 5038.99,/? < .001, o>2 = .89, through aged adult, respectively. Multiplying but this is of no psychological interest because these weights by the number of photos as- photos were deliberately selected to vary signed to each category, and dividing by 5 widely in age. Since this variable, however, (the number of categories) yields an index captures the major portion of the variance, the of skew in which larger values reflect more power of effects exerted by the other indefrequent use of the older age categories in the pendent variables is severely curtailed. The assignment of stimulus persons. low co2 values reported below must be inFig. 1 presents the relevant data. It can be terpreted within the foregoing context. seen that there was heavier use of the older Consistent with earlier research (Kogan, age categories for female in contrast with male 1975), age estimates were significantly associated with subject age, F (4, 140) = 9.64, p < .001, a)2 = .01. Mean age assignments were ! femde photos as follows: 44.6, 45.0, 44.7, 45.7, and 48.5 for male photos the age groups 18-21, 22-28, 29-38, 39-55, and 56-76, respectively. Application of the Scheff6 test to these data indicated that the oldest subjects (those 56 to 76 years of age) differed significantly from all of the other age groups, whereas the latter did not differ from one another. Females' mean age estimates were somewhat higher than those of males (46.3 vs 45.2), F (1, 140) = 5.46, p < .03, co2 = .002. Male photos elicited higher chronological age estimates than did female photos (46.2 vs 45.2), F ( l , 140) = 13.05,/? < .001, o>2 = .001. This latter finding is of particular interest in the 22-28 29-38 39-55 light of the earlier evidence cited indicating AGE OF SUBJECT that female photos tend to be assigned to older Fig. 1. Skew in age category assignment as a function age categories. This implies that the perceived transition from one age category to the next of sex of photo and age of subject.

KOGAN

362

subjects: • 08 //, • iH

17-21 yrs. 22-28 yrs. 29-38yrs. 39- 55 yrs. 56-76yrs.

female photos mde photos

35

25

25

AGE CATEGORY ASSIGNMENT OF PHOTOS

Fig. 2. Mean estimated age of photos for each age category and subject age group.

Table 1. Distance between Age Categories as a Function of Age of Subject.

YA

MA

EA

AGE CATEGORY ASSIGNMENT OF PHOTOS

Fig. 3. Mean estimated age of photos for each age category and sex of photo.

from 15.7 to 16.9 years. In contrast, the distance between middle-aged and elderly adulthood varied widely across age groups. It was 16.9 19.0 18-21 19.0 years in the youngest age group and 16.1 18.1 22-28 16.0 16.7 29-38 declined steadily to 14.6 years in the oldest 15.7 16.2 39-55 age group. With increasing subject age, in 16.6 14.6 56-76 other words, middle age and elderliness were placed in closer proximity to each other. Fig. 3 displays the significant interaction, occurred at a somewhat younger chronological F(2, 280) = 9.40, p < .001, co2 = .001, between age for women than for men. the sex of the photo and the age category to The significant interaction, F (8, 280) = which it was assigned. For stimulus persons 4.05, p < .001, w2 = .003 between age of categorized as young or elderly adults, the subject and the age category assignment of sex of the photo had little impact on chronophotos is displayed in Fig. 2. For photos as- logical age estimation. Such was not the case signed to the young and middle-aged adult for the middle-aged category, where we obcategories, there was an evident progression in serve that the mean estimated chronological which mean age estimates increased with sub- age of females was about two years younger ject age. The judgments of the oldest age group than that for male photos. were most discrepant. No such monotonic Thus, in the categorizations of our subtrend was obtained for age estimates of jects, middle-age was associated with higher photos assigned to the elderly adult category. chronological age levels in the case of males Here the effect was curvilinear. relative to females. Consistent with this obThere is another feature of these findings servation are the results of a comparable that is not immediately apparent from Fig. 2. analysis based on the age of the youngest As Table 1 shows, the mean distance in esti- photo assigned to the middle-aged category. mated chronological age from young to middle- For all practical purposes, this can be conaged adulthood was fairly constant across sub- sidered the age of onset for middle age. Again ject age groups. The discrepancies ranged a significant interaction, F (2, 280) = 5.94, Age of Subject

Age Distance between Category Means

Young-Middle age

Middle age-Elderly

Downloaded from http://geronj.oxfordjournals.org/ at Indiana Univ. School of Medicine on September 2, 2012

30

AGE CATEGORIZATION

363

subjects'

femcfe photos

I^H moles

male photos

25

888888 i i u - f - i WOQKS r c f n a w s

40 45

20

M,A

EA

AGE CATEGORY ASSIGNMENT OF PHOTOS

Fig. 4. Mean onset age (lowest estimate) of photos for each age category and sex of photo.

p < .01, co2 = .001, was obtained as shown in Fig. 4. There we observe that both middle age and elderliness were perceived to have earlier onsets for females than for males. In sum, there is an evident sex-linked bias in age categorization. Entry into middle-aged and elderly adulthood was perceived to occur at a higher age for males in comparison with females. Females, in other words, were viewed as making an earlier transition from young to middle-aged and from middle-aged to elderly adulthood. Photo Preferences Recall that respondents selected three male and three female stimulus persons as most liked. Since' chronological age estimates of these persons had been obtained previously, the median preferred age for each photo set could be readily assessed. A number of significant effects were observed for the median age of the preferred photos. Subject sex exerted a major influence; for males, the median age that they had assigned to preferred photos was 39.6, wherea- the corresponding age for female subjects was 45.3. This is a highly significant difference, F (1, 140) = 14.88, p < .001, cu2 = .05. A significant effect for sex of photo was also obtained, with the median age of preferred photos somewhat

SEX OF PHOTO

Fig. 5. Median estimated age of preferred photos for male and female subjects and photos.

higher for male than for female stimulus persons (44.0 vs 40.9), F ( l , 140) = 5.90, p < .02, w2 = .02. In addition to these significant main effects for sex of subject and sex of photo, there was a highly significant interaction between them, F ( l , 140) = 12.14,/? < .001, co2 = .03. This is clearly evident in Fig. 5. Note that for male photos, the median preferred age was essentially the same for male and female subjects. For female photos, on the other hand, there was a considerable discrepancy in the median preferred age of male and female subjects. It can be seen that the median age of photos preferred by males was more than ten years younger than the corresponding preference of females. To phrase these results in a somewhat different way, female respondents preferred female and male photos of essentially similar age. In contrast, male subjects exhibited a marked sex discrimination in their preferences; the median age of preferred female stimulus persons was about eight years younger than the median age of preferred males. Sex of the photo also interacted with the age of the subject in influencing photo preferences. Fig. 6 demonstrates this significant

Downloaded from http://geronj.oxfordjournals.org/ at Indiana Univ. School of Medicine on September 2, 2012

35 YA

KOGAN

364

; female photos I male photos

0 P. 40

22-28 ACX OF SUBJECT

Fig. 6. Median estimated age of preferred photos for each age group on male and female photos.

3S

o 5" 30

25

M SEX OF SUBJECT

SEX OF PHOTO

Fig. 7. Age range of preferred photos for male and female subjects and photos.

interaction, F(4,140) = 2.82,/? < .03, co2 = .03. For the youngest subjects, no sex-of-photo difference was observed. For the three middle groups, there was a marked trend favoring preference of younger female and older male stimulus persons. The direction of this difference was reversed in the oldest age group, where we note that the median age for preferred females was slightly higher than that

Similarity Judgment Consider, finally, the similarity judgment data. Recall that subjects for each photo set were requested to specify the two stimulus persons perceived to be most similar. With chronological age estimations available for each photo, it was possible to examine the perceived difference in age between the two stimulus persons deemed most similar. Sex of subject was the only independent variable influencing the magnitude of the age difference between the two photos judged most similar, F ( l , 140) = 7.01,/? < .01, co2 = .03. For male respondents, the two photos deemed most similar were closer in age (a difference of 6.75 years) than was the case for female respondents (a difference of 9.86 years). This finding suggests that age may be a more salient construct for males than for females when judging other people. DISCUSSION

When asked their generalized views about the chronological age parameters of such categories as young, middle-aged, and old adult, respondents readily produced estimates suggesting sharply demarcated age categories. Without exception, clear age gaps were established with little indication of overlap. Such outcomes are no surprise; they follow

Downloaded from http://geronj.oxfordjournals.org/ at Indiana Univ. School of Medicine on September 2, 2012

35

for preferred males. In sum, the bias toward selecting younger females and somewhat older males as potential acquaintances was particularly noticeable in young and middleaged adults. No such bias was observed in the youngest and oldest age groups included in this study. Thus far, we have considered only a median index of photo preference. Such an index does not take into account the range in age among preferred photos. Fig. 7 offers data relevant to this issue. It can be seen that both the sex of the subject and the sex of the photo influenced the age range of preferred photos. Sex of subject exerted the more powerful effect, F( 1,140)= 10.91,/? < .01, G>2 = .04. The perceived age span of the preferred photos was broader for females than males. Also indicated in Fig. 7 is a significant trend, F(\, 140) = 6.08, p < .02, a)2 = .01 for preferred female photos to span a broader range than did preferred male photos.

A GE CA TEGORIZA TION

described. Further research is also called for in connection with the procedure in which age categorization precedes chronological age estimation. If the memory and informationprocessing burden were to be increased by breaking up the age category photo groupings of subjects prior to chronological age estimations, would one continue to find the sharp age demarcations described (Footnote 3)? There is an evident need for an experiment investigating the order effect for age categorization and age estimation where the two orders make comparable cognitive demands. One of the more salient aspects of the present findings is the important impact of the sex of the stimulus person on age categorization and preference. Females were perceived to age somewhat more quickly than do males. The most preferred stimulus persons tended to be younger when female than when male, though this effect was largely confined to male subjects. There is strong reason to believe on the basis of the present data that age is a more salient and value-laden dimension for males than females. Males were more likely than females to use age as a basis for similarity categorization, and males' preferences clearly revealed a bias in the direction of youth. Females appeared to be more relaxed and tolerant about age differences, and made no distinctions between males and females on an agelinked basis. None of this can be considered surprising, of course. The observed differences reflect dominant values in our culture concerning sex and age and their interaction (Beauvoir, 1972). Indeed, Sontag (1972) has coined the term "double standard of aging." Subject age also influences categorization and preference. The more frequent assignment of female stimulus persons to the older age categories is mitigated as subjects become older. Similarly, the general preference for younger female than male stimulus persons is actually reversed in the oldest age group. As individuals become older, in other words, the salience of the stimulus person's sex in categorization and preferences is very much reduced. " How do the outcomes of the present research compare with the earlier work on generalized beliefs about age stage onsets? Here the comparison with Drevenstedt (1976) seems most appropriate given the recency and comprehensiveness of that study. Point-by-

Downloaded from http://geronj.oxfordjournals.org/ at Indiana Univ. School of Medicine on September 2, 2012

naturally from our knowledge of the human categorization process (Bruner et al., 1956). Categorization serves to make both our physical and personal world more simple and manageable. Social psychologists have begun to study the categorization process with particular emphasis on the link to stereotyping (Eiser & Stroebe, 1972), and have shown how the labelling of groups of people results in their being perceived as more similar than would otherwise be the case (e.g., Allen & Wilder, 1975; Billig & Tajfel, 1973; Hamilton, 1976; Tajfel et al., 1971). Following from this work, we should expect that the labelling of people as young, middle-aged, or old would lead respondents to separate them sharply on other attributes as well. The labelled groups should be perceived as decidedly different in chronological age as well as in a host of other characteristics. A major purpose of the present paper has been to show that categorization outcomes are a reflection of the methods that we use to study it. The method employed in the present investigation (where respondents made age estimates prior to categorization) has demonstrated that the boundaries between age categories are not sharply demarcated. Respondents had a general idea of the chronological age range associated with various age categories; there was no confusion in regard to the difference between a young and an elderly adult. Rather, it was the boundaries between adjacent age categories that constituted a gray area. Thus, in the case of the transition period between middle-age and elderliness, placement on one or the other side of that boundary was only weakly connected to the chronological age of the person being judged. In short, categorizations of people will be "tighter" or "looser" dependent upon the attributes at issue. Thus, sex categorization is relatively "tight," whereas age categorization is relatively "loose." One can maintain, of course, that the "looseness" of the age category boundaries is partially a reflection of the unreliability of chronological age estimates across a set of 33 stimulus photos. There is little doubt that a repetition of these estimates would not yield ages that perfectly matched the initial judgments. It would be of interest to observe whether an averaged age estimate over two or more repeated measurements would alter the pattern of outcomes

365

366

KOGAN

mental methods that encourage subjects to categorize on the basis of age. As we have seen, initial categorization of stimulus persons into age classes was followed by chronological age estimates that preserved intraclass homogeneity and impermeable age boundaries. We have, in essence, established the preconditions for stereotyping. In striking contrast, instructions to categorize after age estimation pointed up the relative fuzziness of boundaries between adjacent age stages. It should be noted, however, that any experimental instruction to categorize persons by age necessarily imposes that attribute upon numerous subjects who, if given the opportunity to group photos on attributes of their own choosing, would not spontaneously do so on the basis of age. Supportive evidence for this observation can be found in a study by Kogan (1961). Subjects were given several sets of photo triads varying widely in age and asked to select the pair within each triad considered most similar and to state the basis of the similarity. Across various conditions, approximately one-eighth to one-third of the similarity pairings were formed on the basis of age. In a free-sorting procedure, on the other hand, approximately 40 to 50% of the photo groupings were agebased (Kogan, 1974). In sum, we must acknowledge that the designs employed by psychologists to study age stereotyping may in fact be causing it or, at least, enhancing it. With such knowledge, it should prove possible to design studies in the future that would disentangle the influence of the investigator's methods and the subject's cognitions in the stereotyping process. SUMMARY

Numerous studies have asked respondents to specify the chronological ages associated with particular age stages or categories (e.g., youth, middle-age, old age). The present research inquired into this issue within the context of photo sorting. A set of 33 male and 33 female photos were provided to male and female subjects (divided into five age groups) for chronological age estimation and categorization as adolescent, young, middle-aged, elderly, or aged adult. In addition, subjects chose the three best-liked photos and the two most similar photos within both male and female sets.

Downloaded from http://geronj.oxfordjournals.org/ at Indiana Univ. School of Medicine on September 2, 2012

point comparisons are not possible, of course, for the latter research was based exclusively upon young and old respondents. Nevertheless, certain common trends can be detected in the two sets of data. In both cases, the older respondents provided high chronological age estimates for various age stages. The effect was somewhat vitiated for the "elderly adult" category in the present research, but this may have been caused by the availability of the "aged" category for classifying photos. On the whole, however, we seem to have a pattern of results here that generalizes across two quite different contexts — belief statements and age estimation and sorting of photos. Further consistency is observed in regard to sex of stimulus. In both assessment contexts, females are perceived to enter particular age stages somewhat sooner than do males. On the other hand, comparable subject sex effects were not found; the present research, unlike Drevenstedt (1976), did not observe that females delayed the onset of age stages in comparison with males. Of further interest is the comparability of absolute age estimates associated with age stages. Regrettably such comparisons are not legitimate, for Drevenstedt (1976) used three age stages — young, middle-aged, old — whereas the present work is based upon five. The availability of "adolescence" and "aged" categories could be expected to elevate the age of photos assigned to the "young adult" category and to reduce the age of photos assigned to the "elderly" category. The general pattern of the findings across the two studies confirms the foregoing expectation. There is good reason to believe, then, that the particular age categories employed (whether more coarse or differentiated) will influence the magnitude of chronological age estimates associated with those categories. This is a somewhat speculative conclusion, however, that is in need of further empirical research. For many years, stereotyping has been considered a motivational process whose outcome is the devaluation of outgroups in comparison with one's

A study of age categorization.

Journal of Gerontology 1979 Vol. 34. No. 3, 358-367 A Study of Age Categorization1 Nathan Kogan, PhD2 UMEROUS investigators have used questionnaire...
1MB Sizes 0 Downloads 0 Views